Partitionnement de tracés manuscrits en ligne par modèles markoviens

Abstract : We present an unsupervised approach to cluster sequences. This method is inspired by topology learning methods for hidden Markov models, and is built upon the definition of a distance between Markov models. This type of technique may be used to learn Markovian character models from data or to identify allographs or handwriting styles.


http://archivesic.ccsd.cnrs.fr/sic_00001219
Contributor : Roger T. Pédauque <>
Submitted on : Tuesday, December 7, 2004 - 2:00:51 PM
Last modification on : Tuesday, December 7, 2004 - 2:00:51 PM

Identifiers

  • HAL Id : sic_00001219, version 1

Collections

SHS

Citation

Henri Binsztok, Thierry Artières, Patrick Gallinari. Partitionnement de tracés manuscrits en ligne par modèles markoviens. Jun 2004. <sic_00001219>

Export

Share

Metrics

Consultation de la notice

119